computational fluid dynamics (cfd) to simulate slug flow ... · in the present work, the impact...
TRANSCRIPT
Computational Fluid Dynamics (CFD) To Simulate Slug Flow in
Horizontal Pipeline and Annular Pipe
by
Hassn Hadia
A thesis submitted to the School of Graduate Studies in partial fulfillment of the requirements for
The degree of
Masters of Engineering
Faculty of Engineering and Applied Science
Memorial University of Newfoundland
May, 2018
St. John's Newfoundland
i
Abstract
Slug flow can emerge as a factor in several industrial processes, especially in the oil and gas (O&G)
industry. However, because of the complications that are inherent in multiphase flow, finding or
developing a viable analysis tool has thus far proven elusive. For the past few decades,
computational fluid dynamics (CFD) has served as the preferred approach in the flow analysis of
single phase flow, yet it is only now beginning to be used in multiphase flow. Therefore, if CFD
is to be adopted on a larger scale in the (O&G) industry, it is imperative that we first explore the
wide variety of models currently existing in the commercial realm.
This thesis investigates the commercial CFD package ANSYS 16.2 analysis of (air-water slug
flow) and (water-sand slurry flow) inside a horizontal pipe (2-15 m long with a 0.05-0.059 m
internal diameter) and an annular pipe (2- 4.5 m long, 0.02- 0.088 m inner and 0.04-0.12 m outer
diameter). A range of two-phase air/water flow simulations is carried out using the Eulerian model
with the Reynolds Stress Model (RSM), and the volume of fluid (VOF) model with the Shear-
Stress-Transport (SST) model option of turbulence. The aim is to simulate a range of fluid
velocities between 1.66 and 7 m/s and a range of gas velocities between 0.55 and 11 m/s.
Additional investigations include comparing CFD predictions along with experimental
measurements in the literature and performing sensitivity studies.
In the present work, the impact from fluid-structure interaction (FSI) is demonstrated by using a
2-way coupling in ANSYS, effectively joining CFD and structural analysis. At the same time,
stress and pressure changes are measured, along with the deformational structural response arising
from unsteady multiphase flow. It is hoped that the outcome of this study will assist engineers and
ii
researchers in making better decisions in terms of operation, design, and sizing of two-phase flow
systems, as these systems have broad and promising applications in subsea (O&G) pipelines.
iii
Acknowledgments.
First of all, I wish to thank my supervisor, Dr. John Shirokoff, along with my co-supervisor, Dr.
Aziz Rahman, whose advice and encouragement guided me into the field of O&G research. Both
have had an enormous impact not only on my future career options but also on my life. Under your
direction, I discovered the fascinating depths of CFD multi-phase slug flow in pipes. Also gave
me the necessary technical support. Without your help, I would not have been able to bring my
research to a successful conclusion, and for that I will be thankful to you for the rest of my life.
As well, I extend my gratitude to Mr. Rasel Sultan. Your ongoing encouragement, together with
your deep knowledge and technical support, were crucial in helping me create my model.
I also extend my thanks to the Libyan Ministry of Higher Education through the Canadian Bureau
for International Education (CBIE). Without their financial assistance, this work would not have
proceeded, and for that I am deeply grateful.
Last but not least, I wish to thank my parents and my family for their ongoing encouragement
during the time at Memorial University, and for their life-long love and support, and my wife and
children – thank you so much for everything you do for me. My wife has helped make these past
two years not only endurable but happy.
iv
Table of Contents
Abstract ............................................................................................................................................ i
Acknowledgments.......................................................................................................................... iii
List of Tables ............................................................................................................................... viii
List of figures ................................................................................................................................. ix
Nomenclature ................................................................................................................................. xi
Chapter 1 ......................................................................................................................................... 1
1.1 Introduction ........................................................................................................................... 1
1.2 Objective ............................................................................................................................... 2
1.3 Organization of the Thesis .................................................................................................... 3
Chapter 2 A Computational Fluid Dynamics Study of Two-Phase Slurry and Slug Flow in
Horizontal Pipelines ........................................................................................................................ 4
2.1 Introduction ........................................................................................................................... 5
2.2 Mathematical models ............................................................................................................ 8
2.2.1 Multi-phase Model ......................................................................................................... 8
2.2.1.1 Volume Fractions ..................................................................................................... 8
2.2.1.2 Conservation Equations ........................................................................................... 9
2.2.1.2.1 Continuity Equation .......................................................................................... 9
2.2.1.2.2 Fluid-Fluid Momentum Equations .................................................................. 10
v
2.2.1.2.3 Fluid-Solid Momentum Equations .................................................................. 10
2.2.2 Solids Pressure .............................................................................................................. 11
2.2.3 Solids Shear Stress........................................................................................................ 12
2.3 The Reynolds Stress Model (RSM) .................................................................................... 12
2.3.1 Reynolds Stress Transport Equations ........................................................................... 13
2.3.2 Modeling Turbulent Diffusive Transport ..................................................................... 14
2.3.3 Linear Pressure-Strain Model ....................................................................................... 14
2.4. Numerical method .............................................................................................................. 16
2.4.1 Conservation Equations in Solid Mechanics ................................................................ 16
2.4.2 Elasticity Equations ...................................................................................................... 16
2.4.3 Fluid Structure Interaction ............................................................................................ 17
2.4.4 Flow-induced Vibration ................................................................................................ 17
2.5. Methodology ...................................................................................................................... 18
2.5.1 Geometry and Mesh...................................................................................................... 18
2.5.2 Boundary Conditions .................................................................................................... 19
2.6. Results and Simulation ....................................................................................................... 20
2.6.1 Pressure Drop ............................................................................................................... 20
2.6.2 Pressure Gradient .......................................................................................................... 21
2.6.3. Sensitivity Analysis ..................................................................................................... 22
2.6.3.1 Solid Concentration Contours ................................................................................ 22
vi
2.6.3.2 Profiles of Local Solid Concentration.................................................................... 23
2.6.3.3 Slug Body Length .................................................................................................. 25
2.6.3.4 Pressure Drop ......................................................................................................... 26
2.6.4. Fluid Structure Interaction (FSI) ................................................................................. 27
2.6.4.1 Stress and Deformation .......................................................................................... 27
2.6.4.2 Profile of (FSI) ....................................................................................................... 28
2.7 Conclusions ......................................................................................................................... 29
Chapter 3 Analyses of Slug flow Through Annular Pipe ............................................................. 31
3.1 Introduction ......................................................................................................................... 32
3.2. Mathematical modeling ...................................................................................................... 34
3.2.1 The governing equations .............................................................................................. 35
3.2.1.1 Continuity Equations ............................................................................................. 35
3.2.1.2 Momentum Equation ............................................................................................. 35
3.2.2 The Volume Fraction Equation .................................................................................... 36
3.2.4. The Shear-Stress Transport (SST) 𝐤 −𝝎 Model ....................................................... 36
3.2.4.1 Transport Equations for the SST 𝒌 − 𝝎 Model ..................................................... 37
3.2.4.2 Modeling the Effective Diffusivity ........................................................................ 37
3.2.2.4.3 Modeling Turbulence Production ....................................................................... 39
3.2.5. Modeling the Turbulence Dissipation ......................................................................... 39
3.2.5.1 Dissipation of 𝒌...................................................................................................... 39
vii
3.2.5.2 Dissipation of 𝝎 ..................................................................................................... 40
3.3. Methodology ...................................................................................................................... 40
3.3.1 Geometry and mesh ...................................................................................................... 40
3.3.2 Mesh Independent......................................................................................................... 42
3.3.3 Boundary conditions ..................................................................................................... 43
3.3.4 Convergence Criteria .................................................................................................... 44
3.4. Results and Discussion ....................................................................................................... 44
3.4.1 Velocity Profile in Annular Pipe .................................................................................. 44
3.4.2 Velocity Profile in Pipeline .......................................................................................... 45
3.4.3 Pressure Gradient .......................................................................................................... 46
3.4.4 Profile Slug Volume Fraction in Annular Pipe ............................................................. 47
3.4.5 Sensitivity Analysis Comparison between slug flow volume fraction in pipeline and
annular pipe ........................................................................................................................... 49
3.4.5.1 Slug flow volume fraction in pipeline ................................................................... 49
3.4.5.2 Slug flow volume fraction in annular pipe ............................................................ 50
3.4.5.3 Slug Flow Volume fraction Analysis with Time ................................................... 51
Chapter 4 Conclusions and Recommendations for Future Research Work .................................. 53
4.1 Conclusions ......................................................................................................................... 53
4.2 Recommendations for Future Research Work .................................................................... 54
viii
List of Tables
Table 3.1 Different liquid velocities with constant gas velocity used for simulation................... 49
ix
List of figures
Figure 2-1 Computational mesh used for simulation .................................................................... 19
Figure 2-2 Pressure drop obtained from CFD compared to pressure drop measured from data
with Flow velocity (m/s) ............................................................................................................... 20
Figure 2-3 Comparison between CFD pressure gradient and experimental data Nadler and
Mewes (1995a) for air–water flow, D=0.059m with superficial gas velocity (m/s)..................... 21
Figure 2-4 Sand concentration distribution at fully developed flow regime with 3.1 m/s mixture
velocity. Cv = 14%, Cv = 13.3% and Cv 45% ............................................................................. 23
Figure 2-5 Comparison of simulated and measured values of local volumetric concentration of
solid across vertical centerline for particle sizes 0.18 mm. .......................................................... 24
Figure 2-6 Slug length calculation of air-water slug flow ............................................................ 26
Figure 2-7 Pressure drop along the pipe for all Cases. ................................................................. 27
Figure 2-8 Deformation of straight pipeline: (a) Minimum deformation, (b) Maximum
deformation ................................................................................................................................... 28
Figure 2-9 Total Deformation (m) ................................................................................................ 29
Figure 3-1 Meshing of model used in CFD simulation (A) Mesh of Annular Pipe (B) Mesh of
Pipeline ......................................................................................................................................... 42
Figure 3-2 Mesh independence analysis ....................................................................................... 43
Figure 3-3 Comparison of liquid velocity between simulation and experimental data of Nouri and
Whitelaw, 1994. For liquid Velocity 1.3 (m/s), Pressure outlet = 0 (Annular pipe) .................... 45
x
Figure 3-4 Comparison of mean liquid velocity at cross section simulation and experimental data
of Lewis 2002. For gas velocity 0.55 m/s, liquid velocity 1.65 m/s Pressure outlet = 0 (pipeline)
....................................................................................................................................................... 46
Figure 3-5 Pressure gradient obtained from CFD compared to pressure gradients measured from
Experimental data in pipeline and Concentric Annuli .................................................................. 47
Figure 3-6 Concentric annular slug flow in horizontal annuli (A) ............................................... 48
Figure 3-7 Concentric annular slug flow in horizontal annuli (B)................................................ 48
Figure 3-8 Sectional of liquid and gas volume fraction evaluation along the pipe length ........... 50
Figure 3-9 Sectional of liquid and gas volume fraction evaluation along the annular pipe length
....................................................................................................................................................... 50
Figure 3-10 Slug flow trend at different time lapse indicating time in horizontal pipeline.......... 51
Figure 3-11 Slug flow trend at different time lapse indicating time in horizontal annular pipe... 51
xi
Nomenclature
Name Symbol
Density ρ
Instantaneous velocity u
Pressure p
Viscous stress tensor τ
Gravity vector g
Mass Weighted Favre
Flow quantity f
Mass transfer from the pth to the qth phase mpq
Mass transfer from the qth to the pthphase mqp
Phase reference or volume averaged density of the qthphase ρrq
Viscosity µ
Velocity v
Force term F
Interface between the two phases S
Physical density of one phase (q) ρq
Generation of turbulence kinetic energy due to mean velocity gradients Gk
Generation of ω Gw
Effective diffusivity of k Tk
Effective diffusivity of ω Tw
Dissipation of k due to turbulence Yk
Dissipation of ω due to turbulence Yw
xii
User-defined source terms SK and Sw
Turbulent Prandtl numbers for k σk
Turbulent Prandtl numbers for ω σw
Turbulent viscosity μt
Average rate-of-rotation tensor Ωij
Coefficient damps α∗
Turbulent viscosity for blending functions F1and F2
Distance to subsequent surfaces y
Positive component in the cross-diffusion term Dw+
Turbulence kinetic energy production Gk
ω production Gw
Turbulence kinetic energy dissipation Yk
Dissipation of ω Yw
Constant flexural rigidity EI
Pipe’s internal area A
Pipe length L
Pipe radius R
Normalized radial setting for pipe, r R/r
Gas velocity Vg, Vair
Liquid velocity Vl, Vwater
Time in seconds t
Distance between inner and outer wall of the annular pipe
y/ξ
1
Chapter 1
1.1 Introduction
When gas and water move along a pipeline at the same time, differences in density can cause the
two phases to distribute in several different configurations. In general, operating conditions (e.g.,
pipeline angle and phase velocities) determine the phase distribution in pipelines [1]. One type of
flow is “slug flow”, which is unstable and complicated. Despite the rate of liquid and gas flow
staying more or less the same, extreme variations in time can appear in pipeline cross-sections,
phase velocities and pressure, and component mass flow rates. This leads to the destabilization of
heat and mass transfer processes. At the same time, the interruptions in flow due to slug flow can
lead to vibration and pressure drops throughout the length of the pipe, potentially causing damage
to pipe supports and bend as well as pipe corrosion (if there is sand in the flow). Slug flow can
also detrimentally affect equipment used for the separation process. In this case, slug catchers,
which are a type of pre-separation vessel, must be used to collect the slugs, which can occur in
numerous industrial processes, including those related to oil and gas because such large supplies
of oil and gas are used globally every day, even a minor improvement in the efficiency of the
extraction process would have a major effect on overall industry costs. The key here is to source
the appropriate analysis tools that will aid in optimizing multiphase flows for oil and gas
companies [2].
Computational fluid dynamics (CFD), which was developed and has been used over the past 70
years, is an analysis tool commonly applied to subsea equipment. In the 1990s, CFD was used for
single-phase flow calculations, which were made easier by the introduction of commercially
2
available CFD software like ANSYS Fluent Utilizing CFD for slug flow is, at the time of writing
this thesis, still relatively rare, but given the ramp-up in computer resources that is rendering
complex analyses not only possible but simple, CFD is becoming increasingly better known in this
area [3]. The inclusion of slug flow models within the commercial mentioned above is assisting in
the popularization.
Error sources persist in some of the simulations, which is to be expected. However, usage errors
can be problematic. Misapplication of models, along with inaccurate parameters and boundary
conditions can cause severely skewed results. Therefore, considering the preference for using CFD
simulations in engineering projects, it is imperative to gauge the correctness and aptness of
commercial codes, along with the types of models chosen. This can be especially crucial in slug
flow situations, where complicated physical laws and numerical treatment renders the correct
choice of appropriate models not readily apparent [4].
To date, very little investigation has been made into comparing commercial CFD codes. These
codes and models can be created and applied to specific multiphase area, but codes that are well-
suited to one type of commercial area might not be applicable to another. Given the need for
specificity, it might be necessary and indeed useful to compare all available models, with the aim
of building a knowledge base for slug flow simulations that utilize commercial-grade software [5].
1.2 Objective
In this thesis, the main objective is to use numerical simulations and validation in relation to
experimental data as a means to lay a knowledge foundation suitable for defining multi-phase slug
flow CFD processes. To that end, a two range of multi-phase models on ANSYS CFD software is
3
investigated in order to determine their suitability (i.e., their benefits as well as their limitations)
in slug flow applications. Additionally, the aim is to formulate a CFD model that will not only be
viable but also useful in general research into pipelines and/or annular pipe flows in relation to
multi-phase slug fluids. Finally, we aim to determine the validity of our proposed model when
applied in two-phase air-water slug flow in a pipeline, as well as to explain Fluid-Structure
Interaction (FSI).
1.3 Organization of the Thesis
This thesis is written in manuscript format. Outline of each chapter is explained below:
Chapter 1 is a brief introduction of CFD slug flow in pipes with the ANSYS software.
Chapter 2 presents a Computational Fluid Dynamics Study of Two-Phase Slurry and Slug Flow in
Horizontal Pipelines.
Chapter 3 presents Analyses of Slug flow Through Annular Pipe.
Chapter 4 is the overall conclusion of the study and further potential research scope in this area.
4
Chapter 2
A Computational Fluid Dynamics Study of Two-Phase Slurry and Slug Flow in Horizontal
Pipelines
Hassn Hadia1*, Rasel Sultan1, Mohamed Rahman2, John Shirokoff1, Sohrab Zendehboudi1
1Faculty of Engineering and Applied Science, Memorial University of Newfoundland, Canada
2 Faculty of Petroleum Engineering, Texas A&M University at Qatar, Doha, Qatar
Abstract
This paper presents a Computational Fluid Dynamics (CFD) model to simulate two-phase slurry
(water/sand) and slug (water/air) flow systems through utilizing the ANSYS Fluent simulation
package. The CFD model is used to forecast the start and growth of the slug phase as well as its
effect on horizontal vibrations. Eulerian model with Reynolds Stress Model (RSM) turbulence
closure is considered to numerically analyze the slug and slurry flow of mono-dispersed fine
particles at high concentrations. The Eulerian model provides fairly acceptable predictions while
determining the pressure drop and concentration profile for various effluent concentrations and
flow velocities. Furthermore, the optical observations made at the horizontal pipeline flow are used
for validation of 3D simulation results for both air/water and water/sand horizontal flow systems
where the slug and slurry flow conditions are established. The vibration characteristics of gas/
liquid/ solid particles flow patterns in pipelines are also investigated in this work. Keywords:
CFD, FLUENT, Pipeline, Slug, Slurry, Vibrations.
5
2.1 Introduction
The flow in pipeline or annuli are of great importance and widely applied in different industries,
such as chemical process and petroleum industries, pipe line engineering, power plants, biomedical
engineering applications, micro-scale fluid dynamics studies, food processing industries,
geothermal flows and extrusion of molten plastics [6].
Among all other types of flow, water-solid slurry flow and water-air slug flow have become
increasingly popular due to its numerous applications in different industries and enormous focus
of society on reduction in environmental pollution. This type of multiphase flow frequently occurs
in horizontal pipelines and channels [6]. Liquid-Solid two phase slurry flow has been applied to
transport raw materials, waste and sludge which are in solid form [7], beneficiation in extractive
metallurgy and mining plants [8], coal processing plants [9], fluidized beds [10], food and chemical
plants, petroleum industries and many more. Slurry transportation system helps to reduce traffic,
air pollution, noise, accidents along with saving on energy consumption and lesser ecological
disturbance. On the other hand, slug flow is caused by aerated slugs of liquid that flow down a
pipeline at the same velocity as the gas. Many different operations in an oil field can be at the root
of slugging, such as pigging, start-up, blow-down and general transient effects [11].
These problems can occur in the chemical and process industries or in thermo-hydraulic
engineering for nuclear power plants [12], but our focus here is on oil and gas production. In these
pipelines, multiphase slug flows can develop across a broad range of gas and liquid flow rates and
pipe inclinations. Slug initiation, including slug initiation prediction, has been studied by several
6
researchers. In one study, slug initiation prediction is determined by analyzing the stability of a
stratified flow in a pipeline [13].
At the same time, computational fluid dynamics (CFD), which is a programming and computation
method, is also being applied to investigate the behavior of two-phase flows [14]. Another study
looked into slug initiation and growth using a turbulence 𝑘 − 𝜔 model [15]. Unfortunately,
however, the modeling of two-phase flows for studying liquid and gas phases is not only time- and
labor-intensive, but also intensely difficult due to the involvement of advanced physics and
mathematics computations. Typical issues related to slugging are equipment damage, reduced
production, facilities damage, and operational problems with equipment such as pipelines and
separator vessels. Given the wide array of these and other potential problems, it is crucial to have
a firm grasp not only of the slugging operation itself but also of the mechanisms underlying it.
Phase distribution is a key component when designing engineering structures, mainly due to its
impact on the values of parameters like thermal load and pressure drop. It is thus important to
know both the system’s distribution and flow regime. To that end, dual-phase flow maps can aid
in the defining of flow patterns that may occur under different boundary conditions [16].
The main benefit of these mapping tools is that they do not require the user to carry out extensive
and complex numerical calculations. Instead, slug movement can be determined by alterations in
the liquid slugs and gas bubbles flowing at the top of the liquid films, which combine to form slug
units. Slugs moving at a greater velocity than that of average liquid can initiate strong vibrations,
causing damage to equipment in the direction and assemblage centers [17].
Slug frequency, which is defined as the number of slugs flowing past a certain point in a pipeline
within a certain period of time, is an important factor in determining potential operational
7
difficulties such as pipe vibration and instability, fluctuations in wellhead pressure, and flooding
of downstream facilities. Moreover, high slug frequency can cause pipe corrosion [18].
Our study will focus on pipeline vibration caused by unsteady flow, flow directional changes, pipe
diameter, etc., in the petroleum, natural gas and chemical industries. Severe pipeline vibrations
can impact the operation of pipelines and lead to unsafe and even hazardous conditions. Although
pipe vibration is catching the attention of increased numbers of people in the industry, the majority
of investigations into the phenomenon thus far have been on pipe vibration due to mechanical
vibration sources. The cause of fluid vibration in pipelines has been studied with the help of various
theoretical methods. For the sake of simplification and assumption, some results from some these
approaches will be used here as references. Fluid vibration may present in several different forms,
such as gas-liquid flow vibration, high-speed flow vibration, fluid pulsation, and flow vibration
outside the pipeline [19].
Most existing studies focus on flow-induced vibrations (FIV) (impacts on internal flow from
external current), whereas less attention has been allotted to internal flow, slug surge, and external
current.The main aim of the current investigation into issues related to fluid structure interaction
(FSI) is to develop a methodology that explains the basic physics of (FSI), along with the impact
of the phenomenon on subsea piping parts. The accumulated data from this study (as well as studies
in the future) will help to reformulate and revise the FSI model and its capabilities. Potential areas
of improvement are to include Reynolds numbers and to show how free stream turbulent intensity
levels impact subsea piping [20].
8
2.2 Mathematical models
The Eulerian multi-phase model of granular version is used in the current study. Although several
different factors can be involved, the choice of a suitable model relies primarily on the range of
volume fraction 𝛼𝑞of the solid phase under consideration. Hence, given the high value of volume
fraction used here, the granular version emerges as the most appropriate. This approach helps to
indicate the effects of friction and collusion among particles, an ability that is particularly desirable
in high-concentration slurries with different sized grain [21].
2.2.1 Multi-phase Model
The Eulerian multi-phase model enables the modeling of different types of interactive phases, such
as solids, liquids or gases, or any combination of these three states. Unlike the Eulerian-Lagrangian
treatment, which is utilized in discrete phase models, the Eulerian approach is applied to individual
phases [21].
2.2.1.1 Volume Fractions
Multi-phase flow, which can be described as “interpenetrating continua incorporate[ing] the
concept of phasic volume fractions”, is indicated here by 𝛼𝑞. [21]. Volume fractions indicate the
area covered by each phase, while conservation laws pertaining to mass and momentum are
satisfied by the phases. Furthermore, conservation equations can be calculated either by averaging
9
the local balance for each phase or by applying the mixture theory [22]. Phase q, 𝑉𝑞 volume is
stated as:
𝑉𝑞=∫𝛼𝑞 𝑑𝑣 (2-1)
where
∑ = 1𝛼𝑞𝑛𝑞 = 1 (2-2)
Furthermore, phase q’s effective density is calculated as:
𝑞 = 𝛼𝑞𝜌𝑞 (2-3)
with 𝜌𝑞 being phase q’s physical density.
2.2.1.2 Conservation Equations
To designate general instances of n-phase flow, some equations for fluid-fluid and granular multi-
phase flows are given below.
2.2.1.2.1 Continuity Equation
Here, the volume fraction for each phase is given in the following continuity equation:
1
𝜌𝑟𝑞[𝜕
𝜕𝑡(𝛼𝑞𝜌𝑞) + ∇. (𝛼𝑞𝜌𝑞
𝜗𝑞→ ) = ∑ (𝑝𝑞 − 𝑞𝑝)
𝑛𝑞=1 ] (2-4)
where 𝜌𝑟𝑞 denotes the phase reference or volume averaged density of the 𝑞𝑡ℎphase and
10
the solution domain, respectively, 𝑝𝑞 characterizes the mass transfer from the 𝑝𝑡ℎ to the 𝑞𝑡ℎ
phase, and 𝑞𝑝 characterizes the mass transfer from the 𝑝𝑡ℎ to the 𝑞𝑡ℎ phase.
2.2.1.2.2 Fluid-Fluid Momentum Equations
Conservation of momentum for the fluid phase, q, is calculated as:
𝜕
𝜕𝑡(𝑎𝑞𝜌𝑞𝜗𝑞 ) + ∇. (𝑎𝑞𝜌𝑞𝜗𝑞 𝜗𝑞 ) = −𝑎𝑞∇𝑃 + ∇. 𝜏𝑞 + 𝑎𝑞𝜌𝑞𝑔 + ∑ 𝑘𝑝𝑞(𝜗𝑝 − 𝜗𝑞 ) + 𝑝𝑞𝜗𝑞𝑝 −
𝑛𝑞=1
𝑞𝑝𝜗𝑞𝑝 + (𝐹𝑞 + 𝐹 𝑙𝑖𝑓𝑡,𝑞 + 𝐹 𝑣𝑚,𝑞) (2-5)
where 𝑔 denotes gravity-driven acceleration, 𝜏𝑞 denotes the 𝑞𝑡ℎ phase stress-strain tensor, 𝐹𝑞
denotes an external body force, 𝐹 𝑙𝑖𝑓𝑡,𝑞 indicates lift force, and 𝐹 𝑣𝑚,𝑞 is virtual mass force.
2.2.1.2.3 Fluid-Solid Momentum Equations
In 1960, Alder and Wainwright [23]. published a research study which presented a multi-fluid
granular model. Their work is used in this present study to describe the flow behavior of a fluid-
solid mixture. As can be seen, the conservation of momentum for the fluid phases is similar to
Equation (2-5). The 𝑠𝑡ℎ solid phase is calculated as follows:
𝜕
𝜕𝑡(𝑎𝑠𝜌𝑠𝜗𝑠 ) + 𝛻. (𝑎𝑠𝜌𝑠𝜗𝑠 𝜗𝑠 ) = −𝑎𝑠𝛻𝑃 − ∇𝑝𝑠 + 𝛻. 𝜏𝑠 + 𝑎𝑠𝜌𝑠𝑔 + ∑ 𝑘𝑙𝑠(𝜗𝑙 − 𝜗𝑠 ) +
𝑛𝑞=1
𝑙𝑠𝜗𝑙𝑠 − 𝑠𝑙𝜗𝑠𝑙 + (𝐹𝑠 + 𝐹 𝑙𝑖𝑓𝑡,𝑠 + 𝐹 𝑣𝑚,𝑠) (2-6)
where 𝑝𝑠 is the 𝑠𝑡ℎ solids pressure, 𝑘𝑙𝑠= 𝑘𝑠𝑙 is the momentum exchange coefficient between
fluid and solid phase 𝑙 and solid phase s, and n denotes the number of phases.
11
2.2.2 Solids Pressure
When granular flows are in the compressible regime (such as when the solid’s volume fraction is
lower than the maximum allowed value), a solid’s pressure can be measured individually and then
substituted for 𝛻𝑝𝑠, which is the pressure gradient term from the granular-phase momentum
equation. Moreover, given that a Maxwellian velocity distribution is applied to the particles, the
factor of granular temperature is thus included in the model:
𝑝𝑠 = 𝑎𝑠𝜌𝑠𝜃𝑠 + 2𝜌𝑠(1 + 𝑒𝑠𝑠)𝑎𝑠2𝑔𝑜,𝑠𝑠𝜃𝑠 (2-7)
where 𝑒𝑠𝑠 denotes the restitution coefficient of particle collisions, 𝑔𝑜,𝑠𝑠 is the radial distribution
function, and 𝜃𝑠 the granular temperature. A default value of 0.9 for 𝑒𝑠𝑠 is applied; however, this
can be changed according to particle type. Furthermore, the granular temperature, 𝜃𝑠 , is shown to
be proportional to the fluctuating particle motion’s kinetic energy, while the function 𝑔𝑜,𝑠𝑠 presents
as a distribution function which determines the steady alteration from the compressible state of
𝑎 < 𝑎𝑠,𝑚𝑎𝑥, such that the area between the solid particles demotes to an incompressible state [23].
Here, 𝑎 = 𝑎𝑠,𝑚𝑎𝑥,, indicating that no additional decreases in area are possible. Although the
default for 𝑎𝑠,𝑚𝑎𝑥, is the value of 0.63, this can change during the process of setting up the problem
[24].
12
2.2.3 Solids Shear Stress
Particle momentum exchange due to translation and collision can lead to solids stress tensor
containing shear and bulk viscosities. Moreover, viscosity, as a frictional component, may
contribute to the viscous-plastic transition which can occur if solid-phase particles achieve a
maximum solid volume fraction. To give the solids shear viscosity ( 𝜇𝑠), we can add collisional
and kinetic parts, as well as an optional frictional part, as shown in the following equation:
𝜇𝑠 = 𝜇𝑠,𝑐𝑜𝑙 + 𝜇𝑠,𝑘𝑖𝑛 + 𝜇𝑠,𝑓𝑟 (2-8)
where 𝜇𝑠,𝑐𝑜𝑙 indicates shear viscosity due to collision, 𝜇𝑆,𝑘𝑖𝑛 denotes kinetic viscosity, and 𝜇𝑠,𝑓𝑟 is
frictional viscosity. The shear viscosity’s collisional portion can thus be modeled as:
𝜇𝑠,𝑐𝑜𝑙 =4
5𝑎𝑠𝜌𝑠𝑑𝑠𝑔𝑜,𝑠𝑠(1 + 𝑒𝑠𝑠) (
𝜃𝑠
𝜋)
1
2 (2-9)
with the kinetic viscosity default expression stated as:
𝜇𝑠,𝑘𝑖𝑛 =𝑎𝑠𝑑𝑠𝜌𝑠√𝜃𝑠𝜋
6(3−𝑒𝑠𝑠)[1 +
2
5(1 + 𝑒𝑠𝑠)(3𝑒𝑠𝑠 − 1)𝑎𝑠𝑔𝑜,𝑠𝑠] (2-10)
Here, frictional viscosity can be added through the expression:
𝜇𝑠,𝑓𝑟 =𝑝𝑠 sin∅
2√𝑙2𝐷 (2-11)
2.3 The Reynolds Stress Model (RSM)
The Reynolds stress equation model (RSM) is the most complete of the classical turbulence
models. Instead of applying the isotropic eddy-viscosity hypothesis, the RSM resolves the
13
Reynolds-averaged Navier Stokes equations by using transport equations for Reynolds stresses,
along with a dissipation rate equation. Hence, four extra transport equations are needed for the 2D
flows, while seven extra transport equations have to be solved in 3D.
However, because the RSM takes into consideration the impacts of streamline curvature, swirl,
rotation, and rapid changes in strain rate using a more in-depth approach than either one- or two-
equation models, it has a higher likelihood to arrive at more accurate predictions for complex
flows. Nonetheless, the accuracy of this model’s predictions can be affected by closure
assumptions used to show different terms in the Reynolds stresses transport equations [24].
In this regard, pressure-strain and dissipation-rate term modeling is especially difficult, thus
leading to the assumption that these measurements can significantly skew RSM prediction
accuracy. Although the RSM might not consistently give results that are better than simpler models
with regard to all flow classes, the approach is still useful when the targeted flow features result
from anisotropy in the Reynolds stresses, such as cyclone flows, swirling flows in combustors,
rotating flow passages, and stress-induced secondary flows in ducts [25].
2.3.1 Reynolds Stress Transport Equations
The exact form of the Reynolds stress transport equations may be derived by taking moments of
the exact momentum equation.
𝜕
𝜕𝑡(𝜌𝑢𝑙
′𝑢𝑙′)⏟
𝐿𝑜𝑐𝑎𝑙 𝑇𝑖𝑚𝑒 𝐷𝑒𝑟𝑖𝑣𝑎𝑡𝑖𝑣𝑒
+𝜕
𝜕𝑥𝑘(𝜌𝑢𝑘𝑢𝑙
′𝑢𝑙′)
⏟ 𝑐𝑖𝑗≡𝐶𝑜𝑛𝑣𝑒𝑐𝑡𝑖𝑜𝑛
= −𝜕
𝜕𝑥𝑘[𝜌𝑢𝑙
′𝑢𝑙′𝑢𝑘′ + 𝑝(𝑠𝑘𝑗𝑢𝑙
′ + 𝑠𝑙𝑘𝑢𝑙′)]
⏟ 𝐷𝑇,𝑖𝑗≡𝑇𝑢𝑟𝑏𝑢𝑙𝑒𝑛𝑡 𝐷𝑖𝑓𝑓𝑢𝑠𝑖𝑜𝑛
+
𝜕
𝜕𝑥𝑘[𝜇
𝜕𝑦
𝜕𝑥𝑘(𝑢𝑙′𝑢𝑙′)]
⏟ 𝐷𝐿,𝑖𝑗≡𝑀𝑜𝑙𝑒𝑐𝑢𝑙𝑎𝑟 𝐷𝑖𝑓𝑓𝑢𝑠𝑖𝑜𝑛
− 𝜌 (𝑢𝑙′𝑢𝑘′ 𝜕𝑢𝑗
𝜕𝑥𝑘+ 𝑢𝑙
′𝑢𝑘′ 𝜕𝑢𝑗
𝜕𝑥𝑘)
⏟ 𝑃𝑖𝑗≡𝑆𝑡𝑟𝑒𝑠𝑠 𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑜𝑛
−
14
𝜌𝛽(𝑔𝑖𝑢𝑗𝑖𝜃 + 𝑔𝑖𝑢𝑗
𝑖𝜃)⏟ 𝐺𝑖𝑗≡𝐵𝑢𝑜𝑦𝑎𝑛𝑐𝑦 𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑜𝑛
+ 𝑝 (𝜕𝑢𝑙
′
𝜕𝑥𝑗+𝜕𝑢𝑙
′
𝜕𝑥𝑙)
⏟ ∅𝑖𝑗≡𝑃𝑟𝑒𝑠𝑠𝑢𝑟𝑒 𝑆𝑡𝑟𝑎𝑖𝑛
− 2𝜇𝜕𝑢𝑙
′
𝜕𝑥𝑘+𝜕𝑢𝑙
′
𝜕𝑥𝑘
⏟ ∈𝑖𝑗≡𝐷𝑖𝑠𝑠𝑖𝑝𝑎𝑡𝑖𝑜𝑛
−
2𝜌Ω𝑘 (𝑢𝑗′𝑢𝑚′ ∈𝑖𝑘𝑚+ 𝑢𝑙
′𝑢𝑚′ ∈𝑗𝑘𝑚)⏟ 𝐹𝑖𝑗≡𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑏𝑦 𝑆𝑦𝑠𝑡𝑒𝑚 𝑅𝑜𝑡𝑎𝑡𝑖𝑜𝑛
+ 𝑆𝑢𝑠𝑒𝑟⏟𝑈𝑠𝑒𝑟−𝐷𝑒𝑓𝑖𝑛𝑒𝑑 𝑆𝑜𝑢𝑟𝑐𝑒 𝑇𝑒𝑟𝑚
(2-12)
Noteworthy here is that various terms in 𝐺𝑖𝑗 ,𝐷𝐿,𝑖𝑗 , 𝐹𝑖𝑗, and Fij require no modeling, even though
𝐺𝑇,𝑖𝑗, 𝐺𝑖𝑗,∅𝑖𝑗 and ∈𝑖𝑗 need to be modeled to close the equations.
Below, the modeling assumptions that are needed to close the equation set are described in detail.
2.3.2 Modeling Turbulent Diffusive Transport
𝐷𝑇,𝑖𝑗 may be modeled after the generalized gradient-diffusion model
𝐷𝑇,𝑖𝑗 = 𝐶𝑠𝜕
𝜕𝑥𝑘(𝜌
𝑘𝑢𝑘′ 𝑢𝑙′
∈
𝜕𝑢𝑙′𝑢𝑗′
𝜕𝑥𝑙) (2-13)
This equation might, however, lead to numerical instabilities. Hence, we simplified the equation
in FLUENT by applying a scalar turbulent diffusivity:
𝐷𝑇,𝑖𝑗 =𝜕
𝜕𝑥𝑘(𝜌
𝜇𝑡
𝜎𝑘
𝜕𝑢𝑙′𝑢𝑗′
𝜕𝑥𝑘) (2-14)
2.3.3 Linear Pressure-Strain Model
In FLUENT, we model the pressure-strain term, ∅𝑖𝑗 . The typical way to model ∅𝑖𝑗 is to apply the
decomposition as follows:
15
∅𝑖𝑗 = ∅𝑖𝑗,1 + ∅𝑖𝑗,2 + ∅𝑖𝑗,𝑤 (2-15)
where ∅𝑖𝑗,1 denotes the slow pressure-strain term (also called the return-to-isotropy term), ∅𝑖𝑗,2
refers to the rapid pressure-strain term, and ∅𝑖𝑗,𝑤 indicates the wall-reflection term.
As shown, the slow pressure-strain term, ∅𝑖𝑗,1 , can be shown as:
∅𝑖𝑗1 ≡ −𝐶1𝜌𝜖
𝑘[𝑢𝑙′𝑢𝑗′ −
2
3𝛿𝑖𝑗𝐾] (2-16)
where 𝐶1 = 1.8.
Meanwhile, the rapid pressure-strain term, ∅𝑖𝑗,2 , can be modeled as:
∅𝑖𝑗,2≡ 𝐶2 [(𝑃𝑖𝑗 + 𝐹𝑖𝑗 + 𝐺𝑖𝑗 + 𝐶𝑖𝑗 ) −2
3𝛿𝑖𝑗(𝑃 + 𝐺 + 𝐶)] (2-17)
with 𝐶2 = 0.60, 𝑃𝑖𝑗,𝐹𝑖𝑗,𝐶𝑖𝑗 , and 𝐶𝑖𝑗 being defined as previously shown in Equation (2-12), namely
P = 1
2𝑃𝑘𝑘, 𝐺 =
1
2𝐺𝑘𝑘, and 𝐶 =
1
2𝐶𝑘𝑘. The wall-reflection term, ∅𝑖𝑗,𝑤 , indicates the redistribution of
typical stresses close to the wall. Specifically, it dampens the stresses perpendicular to the wall,
but enhances the stresses parallel to it. The equations below model the term:
∅𝑖𝑗,𝑤 ≡ 𝐶1′ ∈
𝑘(𝑢𝑘′ 𝑢𝑚′ 𝑛𝑘𝑛𝑚𝛿𝑖𝑗 −
3
2𝑢𝑙′𝑢𝑘′ 𝑛𝑗𝑛𝑘 −
3
2𝑢𝑙′𝑢𝑘′ 𝑛𝑖𝑛𝑘)
𝑘32
𝐶𝑙∈𝑑
+𝐶2′ (∅𝑘𝑚2𝑛𝑘𝑛𝑚𝛿𝑖𝑗
3
2∅𝑗𝑘,2𝑛𝑗𝑛𝑘 −
3
2∅𝑗𝑘,2𝑛𝑗𝑛𝑘)
𝑘32
𝐶𝑙∈𝑑 (2-18)
where 𝐶1′= 0.5, 𝐶2
′ = 0.3, 𝑛𝑘 is the 𝑥𝑘 component of the unit normal to the wall, d is the normal
distance to the wall, and 𝐶𝑙 =𝐶𝜇34
𝑘, where 𝐶𝜇 = 0.09 and 𝑘 is the von 𝑘𝑟𝑚𝑛 constant (= 0.4187).
Here, ∅𝑖𝑗,𝑤 is added to the Reynolds stress model by default.
16
2.4. Numerical method
2.4.1 Conservation Equations in Solid Mechanics
Solid mechanics is a field of physics that investigates how solids react when impacted by external
loads.
2.4.2 Elasticity Equations
Fluid structure interactions typically lead to elastic or plastic deformations of solid structures,
which is caused by flow-induced forces. Ideally, the material should have an elastic behavior that
enables it to regain its original shape or arrangement following the application of the load.
Although stress can vary linearly, according to strain amount, an elastic deformable solid must
adhere to continuum mechanics. In other words, it must abide by the conservation law that states:
the sum of the forces must equal to zero [26].
The forces cause a distribution of stress throughout the surface area. So, when a large force is
applied, the material might surpass the limitations of the elastic region and thus could fail through
fracturing or assuming plastic behavior. The type of stress to which a material is subjected can
change according to the location where the force is applied. To resolve the issue of stress
components, the most common approach is to apportion the elastic material into smaller elements
[26]. The following calculations are intended for normal and shear stresses:
∂σx
∂x + ∂τxy
∂y + ∂τxz
∂z +xb= 0 (2-19)
17
where 𝜕 denotes normal stress, 𝜏 indicates shear stress, and 𝑥𝑏 represents body forces per unit of
volume.
2.4.3 Fluid Structure Interaction
Fluid Structure Interaction (FSI) is a multi-physics area that studies the impacts of a flow’s
pressure fluctuations on a structure in terms of deformation and stress. It also investigates whether
it is a solid.
2.4.4 Flow-induced Vibration
If pressure fluctuations against the pipe wall are sufficiently large, fluids being transported through
subsea pipes can lead to a phenomenon known as flow-induced vibration (FIV). When dealing
with FIV, the pipe’s instability is heavily dependent on the pipe’s end condition. The type of pipe
most susceptible to FIV damage and failure is a straight pipe with fixed ends. If there is breaching
of the critical velocity, FIV can cause the pipe to buckle, as shown in the following equation:
Vc =π
L(EI
ρA)
1
2 (2-20)
where EI denotes constant flexural rigidity, ρ indicates fluid density, A represents the pipe’s
internal area, and L and is pipe length [26].
18
2.5. Methodology
2.5.1 Geometry and Mesh
In this study, numerical simulations were done using a horizontal pipe (3m in length and 0.05m in
diameter). Given the importance of the mesh to the numerical solution, the material required
specific and exacting characteristics in order to provide a solution that was both feasible and
accurate. The Directed Mesh technique in ANSYS software was used to develop the material and
demonstrated appropriateness for simulating a two phase flow in the horizontal pipe. More
specifically, the Directed Mesh technique was chosen because of its ability to decrease both the
computational time and the number of cells in comparison to alternative meshing techniques, as
well as its ability to form grids parametrically in a multi-block structure. By employing the path
mesh, the user can control and specify the number of divisions in the inlet cross-section, enabling
the creation of quadrilateral faces. Furthermore, by applying a novel type of volume distribution,
users can specify how many layers they want to have on the pipe. To generate volume mesh,
hexahedral grid cells were created through the extrusion of quadrilateral faces along the length of
the pipe at each layer, as shown in Figure 2-1. It was determined that a structured hexahedral grid
was most appropriate in the present case, as such a grid enabled a fine cross-sectional mesh to be
created without also requiring an equivalent longitudinal one. This approach was considered
superior, as it offered a faster process convergence. Additionally, as the fluid domains were
asymmetrical, a grid independency study was carried out based on the water’s superficial velocity
at the outlet. In a multiphase flow, superficial velocity is the ratio of the velocity and the volume
fraction of the considered phase. Hence, actual velocity of phase = (Superficial velocity of phase)/
(volume fraction of phase).
19
Figure 2-1 Computational mesh used for simulation
2.5.2 Boundary Conditions
At the gas and liquid inlets, uniform velocity inlets were used as boundary conditions. As well, an
atmospheric pressure outlet condition was determined for the outlet to prevent any issues related
to backflow at the tube’s outlet, and a no-slip boundary condition was applied at the tube walls.
The effect of the gravitational force on the flow was also taken into consideration. Overall, the
initial volume fraction of gas was altered according to changes in the pipeline's gas velocity
2.5.3 Convergence Criteria
FLUENT is software for simulating flow utilizing pre-stated boundary conditions and a turbulence
model. In order to terminate the A iteration, we use a convergence criterion of 10−6. Furthermore,
to guarantee the desired degree of accuracy as well as stability and convergence of the iterative
process, we use second-order upwind discretization for the momentum equation, along with a first
upwind discretization for volume fraction, turbulent kinetic energy and dissipation.
20
2.6. Results and Simulation
2.6.1 Pressure Drop
In designing a pipeline, pressure is a crucial parameter. Specifically, a system’s pressure readings
are essential measurements in calculating the pumping energy in a flow. In the present study, and
as shown in Figuir 2-2, pressure was obtained from CFD along the pipe between the inlet and
outlet. Similar to work performed by Kaushal, the diameter of the pipe used was 0.054, with a flow
velocity of up to (5 m/s). We found that the pressure drop for single-phase flow rises as the flow
velocity rises. The results indicate that there is very good agreement with the experimental data,
showing < 9% error.
Figure 2-2 Pressure drop obtained from CFD compared to pressure drop measured from data
with Flow velocity (m/s)
0
500
1000
1500
2000
2500
3000
3500
4000
0 1 2 3 4 5
Pre
ssure
dro
p (
Pa/
m)
Flow Velocity (m/s)
Experimental Numerical
21
2.6.2 Pressure Gradient
Figure 2-3 depicts a CFD simulation of Eulerian multiphase model pressure gradients in a two
phase slug flow. As can be seen in the figure, the x axis indicates gas superficial velocity, while
the y axis indicates pressure gradient. Also shown (for comparative purposes) with experimental
data from studies done by Kago et al. and Nadler and Mewes .Using a diameter of 0.05, two sets
of simulations showed experimental results of gas velocities and constant liquid velocity. It is clear
from figure 2-3 that the CFD outcomes demonstrate nearly the exact same trends as those
stemming from the experimental data .The similarities in outcomes thus affirm the suitability of
using CFD to pre-determine pressure gradients in gas and liquid slug flows.
Figure 2-3 Comparison between CFD pressure gradient and experimental data Nadler and
Mewes (1995a) for air–water flow, D=0.059m with superficial gas velocity (m/s)
0
200
400
600
800
1000
1200
1400
1600
1800
2000
0 2 4 6 8 10 12 14
Pre
ssure
Gra
die
nt
(Pa/
m)
Superficial Gas Velocity (m/s)
Numerical VL(0.9 m/s) Numerical VL(0.5m/s)Experimental VL(0.9 m/s) Experimental VL(0.5m/s)
22
2.6.3. Sensitivity Analysis
2.6.3.1 Solid Concentration Contours
Figure 2-4 depict contours in a non-dimensional solid concentration on a vertical plane outlet. As
can be seen, the 4 grain sizes display different efflux concentrations (𝐶𝑣𝑓) with a 3.1m/s mixture
velocity. The vertical plane’s solid concentration at the outlet point is revealed as a non-
dimensional zed when applying the corresponding inlet efflux concentration shown in figure 2-4.
Furthermore, the contours clearly illustrate how areas featuring the highest solid concentrations
are positioned close to the wall at the bottom half of the pipe’s cross-section when pertaining to
small grain (i.e., particle) objects. The positioning of these areas, however, experience a constant
movement upward as the grain particle sizes increase. This movement is likely caused by the rise
in lift-force in objects positioned close to the wall, as has been formulated for larger grain/particle
sized in the simulation tests. Moreover, because anomalies in validation data also occurred close
to the wall areas during testing of larger grain/particle sizes, this indicates a need to model the lift
coefficient employed in simulation tests on the variously sized grains. Additionally, it was noted
how the spread in the solid concentration area was enhanced by boosting the efflux concentration
and mixture velocity; this increase, however, revealed a somewhat reduced intensity.
23
Figure 2-4 Sand concentration distribution at fully developed flow regime with 3.1 m/s mixture
velocity. Cv = 14%, Cv = 29% and Cv 45%
2.6.3.2 Profiles of Local Solid Concentration
Experimental data from Gillies and Shook are compared with a local solid concentration profile
of water-sand slurry flow from a simulation. The length of the pipe used in the experiment is 2.7
m and the diameter is 0.0532 m. The fluid (water) density is 9982 Kg/𝑚3, viscosity 0.001003
Kg/m, while the sand density is 3650 Kg/𝑚3. The wall material is aluminum and features a density
of 2800 kg/m3 and a roughness of 0.2mm. Furthermore, the grain size or mean particle diameter
is 0.18 mm, the mixture velocity is 3.1 m/s, and there are three distinct solid volumetric
concentrations of 14%, 29% and 45%.
24
Figure 2-5 Comparison of simulated and measured values of local volumetric concentration of
solid across vertical centerline for particle sizes 0.18 mm.
Figure 2-5 illustrates a comparison of particle sizes and volumetric concentrations of solid particles
for the pipe radius Y/R. As shwon in the Figure 2-5, the simulated results show good agreement
with the experimental values of grain sizes measuring 0.18 mm, but the simulated values differ
somewhat from the experimental values when in close proximity to the wall, particularly in the
bottom portion of the cross-section. A potential explanation for this occurrence is abrasive
rounding of the large particles due to repeated passages. This could cause fines to be created and
uniformly distributed within the pipe, leading to an increase in carrier density. However, because
data related to this aspect of the experiment were not available in the reference research, the
appropriate adjustments to reflect these data were not made during the simulations.
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6
Pip
e R
adiu
s Y
/R
Volumetric Concentration of Solid Particles
Simulation (14% Vol
Concentration)
Simulation (29% Vol
Concentration)
Simulation (45% Vol
Concentration)
Experiment (14% Vol
Concentration)
Experiment (29% Vol
Concentration)
Experiment (45% Vol
Concentration)
25
The deviations might also have resulted from the value of the static settled concentration (packing
limit) applied in the simulations. Specifically, the 0.63 value used is most suited to calculations
pertaining to very fine grain sizes. To minimize deviations with experimental results, analysis of
newer boundary conditions at the wall for slurry pipeline flows should therefore consider larger
grain sand sizes.
2.6.3.3 Slug Body Length
Figure 2-6 shows simulation test outcomes for 3 different gas velocities, 3.1m/s, 3.5m/s and 4.1m/s
with length of the pipe used 4m and the diameter is 0.051m. As depicted in the figure 2-6, the slug
length stretches along the length of pipe, and there is a proportional relationship for air superficial
velocity and slug length. Specifically, the prediction outcomes for slug length indicate that rises in
air superficial velocity resulted in the lengthening of slugs.
26
Figure 2-6 Slug length calculation of air-water slug flow
2.6.3.4 Pressure Drop
An important guideline for two phase pipeline design is pressure drop, especially with regard to
losses and generated forces interacting with a pipe’s inner surface. In the present work, pressure
was first simulated and then stored in the pipe as a time series by applying ANSYS Software’s
field function. In the simulations, the superficial gas velocities were measured as 3.1, 3.5 and 4.1.
As shown in Figure 2-7, a sudden pressure rebound occurred as a slug impacted the upper pipe
wall, causing a pressure repulsion. This sudden increase in pressure pointed to the slug reaching
the simulation’s pressure sensor, whereas the abrupt pressure decrease pointed to the slug having
passed the sensor. As illustrated in Figure 2-7, the simulation results show how decreases in
pressure within the pipe became more pronounced as the superficial gas velocity increased, even
as the water superficial velocity stayed the same.
27
Figure 2-7 Pressure drop along the pipe for all Cases.
2.6.4. Fluid Structure Interaction (FSI)
2.6.4.1 Stress and Deformation
The fluid-structure interaction (FSI) of a static structural was modelled in ANSYS workbench by
importing the fluid pressure data from Fluent to the static structure domain. The deformation was
checked through a horizontal straight pipe of 3 m (in length) and 0.05 m (in diameter).The stress
and total deformation of the pipeline due to the slug-flow induced vibration was analyzed and
presented in a contour plot, which is presented in Figure 2-8. Growing waves in the gas-liquid
stratified flow in horizontal pipe transformed to a roll waves, which is big enough to seal/bridge
the pipe diameter resulting in air pockets and slugs. The effect of intermittent slug evolution and
flow disturbances in rigid shape results in high pressure gradient. The pressure load to the structure
0
1000
2000
3000
4000
5000
6000
0 0.5 1 1.5 2 2.5 3 3.5 4
Pre
ssure
Dro
p (
pa/
m)
Pipe length (m)
Vg (4.1m/s) Vg (3.5m/s) Vg (3.1m/s)
28
has been demonstrated to cause significant structural deformation in the FSI analysis as presented
in Figure 2-8.
(A)
(B)
Figure 2-8 Deformation of straight pipeline: (A) Minimum deformation, (B) Maximum
deformation
2.6.4.2 Profile of (FSI)
Figure 2-9 represents the total deformation with changing gas velocity. Deformation in pipeline
increases with increasing gas velocity.
29
Figure 2-9 Total Deformation (m)
2.7 Conclusions
This work presented a numerical model aimed at achieving the qualitative study of a horizontal
pipe’s two-phase slug and slurry flows. FLUENT software (computational fluid dynamics (CFD)
package) was used in the investigation. Given the large amount of computational operations that
would have been required, three-dimensional simulation was simply too costly, so a model
simulation was developed based on an Eulerian model and Reynolds stress model (RSM)
turbulence. The results demonstrated all of the phenomena related to slug flow in a 3-D model.
Overall, the comparison of pressure drops in pressure gradient measurements for both single-and
two-phase flows in horizontal pipes indicate good agreement with the experimental data. Hence,
this work adds to the knowledge base around two-phase slurry flows that feature various sized
particles.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0 0.5 1 1.5 2 2.5 3 3.5
Def
orm
atio
n (
m)
Gas Velocity (m/s)
30
However, due to some data scatter related to flow parameters (particularly in slurry flows with
larger particle sizes), the model applied here requires some revisions. High levels of vibration can
be caused by instability arising from two-phase flows (i.e., slug flow), which can then shorten the
pipe’s fatigue life. Because of the complexity of slug flow behavior, engineers have had difficulty
over the years trying to develop a methodology that can anticipate the impact of slugs.
Furthermore, the present work demonstrated that the approach has application in problems related
to fluid structure interaction in the oil and gas industry. Flow-induced vibration (FIV) is a common
occurrence in this industry as a result of the strong coupling of structure and flow. Slug flow can
generate large amplitude deformation such that the structure of the flow is altered and can create
a catastrophic event However, assessing hydraulic characteristics and quantitatively investigating
pipeline slug flow can benefit from more accurate simulations.
31
Chapter 3
Analyses of Slug flow Through Annular Pipe
Hassn Hadia1, Rasel Sultan1, Mohamed Rahman2, John Shirokoff1
1Faculty of Engineering and Applied Science, Memorial University of Newfoundland, Canada
2 Faculty of Petroleum Engineering, Texas A&M University at Qatar, Doha, Qatar
Abstract
Slug flow pattern through straight annular pipe (2- 4.5 m long, 0.02- 0.088 m inner, and 0.04-0.12
m outer diameter) and straight horizontal pipe (2-15 m long and 0.05 m internal diameter), is
discussed in this study. The paper presents and investigates numerical results from computational
fluid dynamic (CFD) simulations of the air-water slug flow where concentric annular pipe
geometry with horizontal orientation is used. The Eulerian model and volume of fluid (VOF)
model with the Shear-Stress-Transport (SST) model option of turbulence closure are used to
simulate slug flow. A commercial CFD package ANSYS 16.2 is used to model slug flow.
Additional investigations include comparing CFD predictions along with experimental
measurements in the literature, and performing sensitivity studies based on different parameters
by changing liquid velocity, gas concentration and timing. Output parameters (such as pressure
gradient; superficial velocity of liquid and gas, air volumetric fraction) are analyzed during the
process. Overall good agreement found in this simulation with experimental data for slug flow in
annular pipe.
Keywords: CFD, FLUENT, VOF, Annular Pipe, Slug flow
32
3.1 Introduction
For companies working in the Oil and Gas (O&G) field, annular pipe slug flow can be a crucial
factor in applications related to drilling equipment [27]. In fact, annular pipe flow can offer an
overview of issues in relation to advanced stages of turbulent slug flow. Concentric annular pipe
flow features dual boundary layers that function to distribute a range of turbulence qualities, with
pipe and channel flows serving as the limiting versions of annular pipe slug flow [28].
Overall, the O&G industry experiences a broad spectrum of issues around the transportation of
fluids (including multiphase flow) either through pipelines or annular pipes. That being said,
multiphase flow domains can now utilize advanced technology in the form of fluid dynamic
software packages that enable engineers to tailor pipe designs to exact specifications in order to
predict the output of O&G systems [29]. Such specifications are useful, given that unstable
pipeline flow can cause operational issues that ultimately increase expenditures. So, for instance,
liquid flows that contain large slugs are considered unstable and generally require a separator to
deal with them; otherwise, they could further evolve into hydrodynamic slugs or even a mass of
slugs [30].
The mechanisms undergirding slug growth are currently not well understood. Typically, slugs are
the end result of disturbances in liquid and gas plugs. Slug flow occurs in liquid-gas two-phase
flows in a horizontal pipeline and annular pipe across a range of intermediate flow rates. The
slug/plug distribution depends on a variety of factors, such as fluctuations in gas/liquid velocity,
pigging, and even the terrain [30]. Each slug/plug unit is made up of a gas bubble and aerated
33
liquid slug. Bubble size depends mainly on positioning, with smaller bubbles occurring
immediately in front of slugs, and larger ones trailing behind [31].
However, unlike the bubbles, slug lengths are consistent even through pressure drops in the
pipeline. Because gas and liquid multiphase flow has become so common in, for instance, O&G
transport as well as other critical industries involving geothermal heating, there has been an
increase in the published literature dedicated to the topic of flows in pipes and tubing with
diameters smaller than 10 cm. However, data pertaining to large diameter pipelines is nearly non-
existent in the literature [32]. Further adding to the problem of data gap is the inconsistency of
slugs in two-phase gas/liquid, which necessitates very complicated simulations involving, among
other measurements, the flow field geometry of flow regimes or flow patterns [33]. Additional
issues can arise when attempting to assess the phase tile distribution, as well as factors like pressure
drop, pressure gradient, and heat/mass transfer [34].
Direct numerical simulations (DNS) solve the main equations without the use of models, but they
can be very time-consuming. This is because all flows can be described as ‘turbulent’, and
turbulent flows contain immensely different magnitudes of time scales. Consequently, calculation
times become unviable due to the fineness of the mesh resolution. In these instances, modeling
can be used to consider any turbulent effects. In fact, turbulence modeling has recently become the
primary focus of single-phase CFD research [35].
While in Ghosh et al. looked at the need for Multiphysics flow field information for economic
design and optimization of operating conditions in so doing, they applied CFD to simulate the air-
water flow situation [36]. Kaushik et al. [37] utilized the CFD software package FLUENT 16.2 to
simulate the annular flow through horizontal pipes and succeeding in matching the simulated data
and experimental results. They also carried out a simulation to investigate annular flow under
34
conditions of sudden expansion and contraction. In analyzing the profiles of volume fraction,
velocity, pressure, and the fouling characteristic, the researchers presented a CFD analysis of core
annular flow through pipeline and talked about the distribution of pressure, velocity and volume
fraction, along with the fouling characteristic. Feasible operation conditions were then suggested,
but these works are simulated by the Shear-Stress-Transport (SST) model and did not mention the
effect of annulus thickness on the annular flow [38]. Ghosh used the Eulerian model to simulate
core annular flow through a pipe and discussed additional flow specifications and the impact of
annulus thickness on core annular flow. As in their results likewise pointed to feasible operational
conditions for pipe design [38].
In this present study, the Eulerian model and Volume of Fluid (VOF) method with the Shear-
Stress-Transport (SST) turbulence model have been implemented, employing the commercial
ANSYS 16.2 software to simulate the horizontal sections of annulus pipe and pipeline for air-water
slug flow. The objective has been to investigate to validate our model with different experimental
data, also to evaluate volume fraction profile with time, in two different cross sections along the
length of the pipes.
3.2. Mathematical modeling
In this work, we apply formulations that indicate the range of fluid-flow. Equations (3-1) and (3-
2) (Navier-Stokes equations) refer to various flow types and can be solved for nearly all flows for
CFD models [39]. We also use (e.g., energy and turbulence equations) to simulate slug:
∂ρ
∂t+∇ ∙ (ρu) = 0 (3-1)
35
∂ρu
∂t+ ∇ ∙ (ρuu) = −∇P + ∇ + ∇ ∙ τ + ρg (3-2)
where 𝜌 denotes density, u denotes instantaneous velocity, p indicates pressure, 𝜏 refers to viscous
stress tensor, and g describes the gravity vector.
3.2.1 The governing equations
Any flow quantity f is split into mean and fluctuating component as f = f + f" with f" = 0 and f=
ρf
ρ .The (f) overbar quantity represents Reynolds averaged mean quantity [40].
3.2.1.1 Continuity Equations
Between phases, interface tracking can be achieved using a continuity equation to calculate the
volume fraction of the phases. In calculating the 𝑞𝑡ℎ phase, the equation can take the form:
1
𝜌𝑟𝑞[𝜕
𝜕𝑡(𝛼𝑞𝜌𝑞) + ∇. (𝛼𝑞𝜌𝑞 𝜗𝑞 ) = ∑ (𝑝𝑞 − 𝑞𝑝)
𝑛𝑞=1 ] (3-3)
where ρrq is denotes the phase reference or volume averaged density of the qthphase and the
solution domain, respectively, mpq characterizes the mass transfer from the pth to the qth phase,
and mqp characterizes the mass transfer from the pth to the qth phase [40].
3.2.1.2 Momentum Equation
As the momentum equation is solved for the entire domain. Moreover, the velocity field is seen to
be identical across all cell phases, despite showing variations between the cells:
36
∂ ∂
∂t(ρv ) + ∇ · (ρv v ) = −∇p + ∇ · [μ (∇v + ∇v T )] + ρg + F (3-4)
In this equation, ρ is the density, v is the velocity, µ is the viscosity, p is the pressure, g is the
gravitational acceleration, and F is the source term. Furthermore, any sizeable velocity difference
between the phases may cause a reduction in the accuracy of the velocity computations closer to
the interface [41].
3.2.2 The Volume Fraction Equation
The tracking of the interface(s) between the phases is accomplished by the solution of a continuity
equation for the volume fraction of one (or more) of the phases. For the 𝑞𝑡ℎphase, this Equation
has the following form:
𝜕𝛼𝑞
𝜕𝑡+ 𝑣 . ∇𝛼𝑞=
𝑆𝛼𝑞
𝜌𝑞 (3-5)
Where 𝑆 is the interface between the two phases and 𝜌𝑞 is the physical density of one phase (𝑞).
By default, the source term on the right-hand side of equation (3-5) is zero, but you can specify a
constant or user-defined mass source for each phase [42]. The volume fraction equation was not
solved for the primary phase; the primary-phase volume fraction will be computed based on the
following constraint:
∑ 𝛼𝑞𝑛𝑞=1 = 1 (3-6)
3.2.4. The Shear-Stress Transport (SST) 𝐤 − 𝝎 Model
This section presents the standard and shear-stress transport (SST) k − ω models. FLUENT also
provides a variation called shear-stress transport (SST) k − ω model, so named because the
37
definition of the turbulent viscosity is modified to account for the transport of the principal
turbulent shear stress. It is this feature that gives SST k − 𝜔 model an advantage in terms of
performance over both the standard k − 𝜔 model and the standard k − 𝜔 model. Other
modifications include the addition of a cross-diffusion term in the 𝜔 equation and a blending
function to ensure that the model equations behave appropriately in both the near-wall and far-
field zones [43].
3.2.4.1 Transport Equations for the SST 𝒌 − 𝝎 Model
The SST k − ω model has a similar form to the standard k − ω model
𝜕
𝜕𝑡(𝜌𝑘) +
𝜕
𝜕𝑥𝑖(𝜌𝑘𝑢𝑖) =
𝜕
𝜕𝑥𝑖(𝑇𝑘
𝜕𝑘
𝜕𝑥𝑖) + 𝐺𝑘 − 𝑌𝑘 + 𝑆𝑘 (3-7)
and
𝜕
𝜕𝑡(𝜌𝑤) +
𝜕
𝜕𝑥𝑖(𝜌𝑤𝑢𝑖) =
𝜕
𝜕𝑥𝑖(𝑇𝑤
𝜕𝑤
𝜕𝑥𝑖) + 𝐺𝑤 − 𝑌𝑤 + 𝑆𝑤 (3-8)
In these equations, Gk represents the generation of turbulence kinetic energy due to mean velocity
gradients, calculated Gw represents the generation of ω, calculated Tk and Tw represent the
effective diffusivity of k and ω, respectively, which are calculated as described below. 𝑌𝑘 and 𝑌𝑤
represent the dissipation of k and ω due to turbulence, calculated Dw represents the cross-diffusion
term, as described below. 𝑆𝑤 and 𝑆𝑤 are user-defined source terms [44].
3.2.4.2 Modeling the Effective Diffusivity
The SST k − ω model, effective diffusivities are:
𝑇𝑘 =𝜇 +𝜇𝑡
𝜎𝑘 (3-9)
38
𝑇𝑤 =𝜇 +𝜇𝑡
𝜎𝑤 (3-10)
where 𝜎𝑘 and 𝜎𝑤 represent turbulent Prandtl numbers for k and ω, respectively. Turbulent
viscosity, 𝜇𝑡, can be expressed as:
𝜇𝑡 =𝜌𝑘
𝑤
1
𝑚𝑎𝑥[1
𝛼∗,Ω𝐹2𝛼1𝑤
] (3-11)
where
Ω≡ √2Ω𝑖𝑗Ω𝑖𝑗 (3-12)
𝜎𝑘=1
𝐹1/𝜎𝑘,1+(1−𝐹1)/𝜎𝑘,2 (3-13)
𝜎𝑘=1
𝐹1/𝜎𝑤,1+(1−𝐹1)/𝜎𝑤,2 (3-14)
Ω𝑖𝑗 denotes the average rate-of-rotation tensor while 𝛼∗ indicates the coefficient damps. The
turbulent viscosity for the blending functions, F1and F1, can be expressed as:
F1𝑡𝑎𝑛ℎ(Φ14) (3-15)
𝛷1= 𝑚𝑖𝑛 [𝑚𝑎𝑥 (√𝑘
0.09𝑤𝑦,500𝜇
𝜌𝑦2𝑤) ,
4𝜌𝑘
𝜎𝑤,2𝐷𝑤+𝑦2] (3-16)
𝐷𝑤+ = 𝑚𝑎𝑥 [2𝜌
1
𝜎𝑤,2
1
𝑤
𝜕𝑘
𝜕𝑥𝑗
𝜕𝑤
𝜕𝑥𝑗, 10−2] (3-17)
F2 = 𝑡𝑎𝑛ℎ(Φ22) (3-18)
𝛷1= 𝑚𝑎𝑥 [2√𝑘
0.09𝑤𝑦,500
𝜌𝑦2𝑤] (3-19)
where 𝑦 indicates the distance to subsequent surfaces and 𝐷𝑤+ denotes the positive component in
the cross-diffusion term [44].
39
3.2.2.4.3 Modeling Turbulence Production
Production of 𝐤
𝐺𝑘 Indicates turbulence kinetic energy production and can thus be formulated similar to the standard 𝑘 −
𝜔 model.
Production of 𝛚
𝐺𝑤 indicates ω production and can be expressed as follows:
𝐺𝑤 =𝛼
𝑣𝑡𝐺𝑘 (3-20)
As can be seen, this expression diverges from the standard k − ω model. The two models calculate
𝛼∞ differently. Whereas in the standard k − ω model, 𝛼∞ denotes a constant (0.52), in the SST
k − ω model, 𝛼∞ appears as:
α∞ = F1α∞,1 + (1 − F1)α∞, 2 (3-21)
where
α∞, 1 =𝛽𝑖,1
𝛽∞∗ -
𝑘2
𝜎𝑤,1√𝛽∞∗ (3-22)
α∞, 2 =𝛽𝑖,2
𝛽∞∗ -
𝑘2
𝜎𝑤,2√𝛽∞∗ (3-23)
where ω denotes 0.41. Equations (3-27) and (3-28) show 𝛽𝑖,1 and 𝛽𝑖,2 , respectively, as follows
[44]:
3.2.5. Modeling the Turbulence Dissipation
3.2.5.1 Dissipation of 𝒌
Yk indicates turbulence kinetic energy dissipation and can be expressed nearly the same way as the
standard k − ω model [44]. The sole difference can be found in how we calculate for fβ∗. Thus,
40
for the standard k − ω model, fβ∗ can be calculated like a piecewise function, whereas for the
SST k − ω model, fβ∗ represents a constant that is equivalent to 1. Based on this,
Yk = ρβ∗kw (3-24)
3.2.5.2 Dissipation of 𝝎
Yw indicates the dissipation of ω . It can be formulated like the standard k − ω model, except for
how βi and fβ are calculated. Whereas, for the standard k − ω model, βi appears as a constant
(0.072), in the SST k − ω model, fβ denotes a constant that is equivalent to (1) [45]. Therefore,
Yk = ρβ𝑤2 (3-25)
Instead of a constant, βi can be expressed as:
βi = F1βi,1 + (1 − F1)βi,2 (3-26)
where
βi,1= 0.075 (3-27)
βi,2= 0.082 (3-28)
while F1 can be formulated using Equation (3-15).
3.3. Methodology
3.3.1 Geometry and mesh
Because the mesh can have a strong effect on the solver convergence and solution for each CFD
simulation, the quality of mesh used should always be relatively high to enable convergence as
41
well as correct proportions for the simulation. In this study, the initial example showed that were
created (i.e., 400,063 elements at the annular pipe but 397,485 at the cross-section of the pipeline).
Additionally, regarding the time resolution, 0.001 was used as a time-step. This approach resolves
the time resolution for measuring instruments employed by (A) and (B). In those instances, mesh
featuring a 400,063 cell annular pipe geometry measuring 2 m in length and with a 0.02 inner, 0.04
outer m diameter (with measurements of the pipeline being 397,485 cell pipeline geometry, 5 m
length/0.05 m diameter) is considered suitable in the inlet flow comprised of liquid and gas
superficial velocities of 0.55 m/s and 1.65 m/s, respectively. However, in this work, we applied
lower velocities. In the two cases, mesh refinement was performed according to the details below
to verify the outcome from the mesh structuring.
Specifically, these simulations have been performed in a four-processor machine, with a
simulation run-time of several hundred hours. This permits movement of the gas phase upwards
(i.e. from the pipe bottom to the pipe top). The run-time can be decreased by adding processors
that run in parallel. It is also worth noting that the strength of computers will likely increase
exponentially soon. Our mesh was the type known as butterfly grid, as shown in Figure 3-1. For
butterfly grip mesh, another mesh (Cartesian mesh) can be added to the central part of a pipe and
used together with a cylindrical pipe. This approach necessitates the use of several blocks;
however, it represents optimal grid quality in relation to mesh density and orthogonality. Although
building this mesh can be more time-consuming, this can be mitigated by employing ANSYS
software.
42
(A) (B)
Figure 3-1 Meshing of model used in CFD simulation (A) Mesh of Annular Pipe (B) Mesh of
Pipeline
3.3.2 Mesh Independent
Mesh domain simplification is required, as we will only be able to resolve a mathematical model
if we assume linearity. In other words, we must make sure the variables targeted for resolution are
linearized for every cell. Such a requirement indicates that mesh made from finer material (which
can be created through specific refinement stages) should be used in parts of the domain featuring
physical properties that are assumed to be highly volatile. However, before developing a mesh
structure with a relatively small number of elements and then performing analysis, we first make
sure that the quality of the mesh as well as the model coverage can be realistically examined. The
aim here is to recreate mesh structures that have a higher number of elements and then to repeat
the analysis and perform a comparison of the results according to properties found in previous
cases. So, for example, when a case features examining an internal flow through one or more
43
channels, we could potentially use pressure drops in critical regions for comparison purposes.
Next, we would continue moving up to several elements that show results in good agreement with
prior results. In this way, any issues that have emerged due to the mesh structure are removed and
the best possible value related to element number can be arrived at to make the calculations faster.
Figure 3-2 shows pressure changes in region Y caused by raising the number of elements. As can
be seen, we would need approximately between 300,000 up to 700,000 elements to perform a study
with valid outcomes.
Figure 3-2 Mesh independence analysis
3.3.3 Boundary conditions
We used uniform velocity inlets to serve as boundary conditions for gas and liquid inlets. To
prevent backflow near the tube’s outlet, we implemented an atmospheric pressure outlet that
included a no-slip boundary condition near the tube walls. As well, we also took into consideration
0
200
400
600
800
1000
1200
1400
1600
1800
2000
0 100000 200000 300000 400000 500000 600000 700000 800000
Pre
ssure
Gra
die
nt
(Pa/
m)
Number of elements
44
the flow’s response to gravitational force flow, and demarcated the initial air volume fraction for
all instances.
3.3.4 Convergence Criteria
FLUENT is software for simulating flow utilizing pre-stated boundary conditions and a turbulence
model. To terminate the iteration, we use a convergence criterion of Error Digit expected.
Furthermore, to guarantee the desired degree of accuracy as well as stability and convergence of
the iterative process, we use second-order upwind discretization for the momentum equation, along
with a first upwind discretization for volume fraction, turbulent kinetic energy and dissipation.
3.4. Results and Discussion
3.4.1 Velocity Profile in Annular Pipe
To look deeper into the curvature impact, we need to compare average axial velocity profiles in
the entire cross-section of annuli for a variety of inner and outer radius ratios. As it can be seen
comparison in Figure 3-3 with Nouri and Whitelaw (1994), the velocity profiles appear to be
asymmetrical with a decided tilt in the direct of the inner wall. This so-called skewness is the result
of maximum axial velocities in locales near the inner wall. Such locales edge closer towards the
inner wall when the radius ratio drops from 0.04 to 0.02. In other words, the maximum velocity
can be found at y= 0.271 m/s distance from the inner wall. Here, y/ξ indicates distance between
inner and outer wall of the annular pipe.
45
Figure 3-3 Comparison of liquid velocity between simulation and experimental data of Nouri and
Whitelaw, 1994. For liquid Velocity 1.3 (m/s), Pressure outlet = 0 (Annular pipe)
3.4.2 Velocity Profile in Pipeline
Figure 3-4 illustrates Lewis’ (2002) data, representing low/ high values for 𝑉𝑔 = 0.55 m/s, with a
fixed value of 𝑉𝑙 =1.65 m/s and a pipe geometry of 0.05m diameter and 15.4m length. As can be
seen, R/r indicates a normalized radial setting for pipe, r. This is calculated near to the vertical
axis, stretching from the center of the pipe towards the probe, where R represents pipe radius. In
this case, – 1.0 and 1.0 indicate the pipe’s bottom and top, respectively. As can further be seen in
Figure 3-4, the average liquid velocity indicates asymmetrical liquid velocity profiles. Here, the
most pronounced velocities can be found near the top of the pipe. Again, the profile illustrated in
Figure 3-4 demonstrates an identical character as the fully-developed turbulent flow profile and
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.2 0.4 0.6 0.8 1
Liq
uid
Vel
oci
ty (
m/s
)
Distance Between Inner And Outer Wall Ofg The Annulur y/ξ
Experimental Numerical
46
includes a transition zone between them. Our simulation indicate that the mean superficial
velocities correspond well to maximum liquid velocity.
Figure 3-4 Comparison of mean liquid velocity at cross section simulation and experimental data
of Lewis 2002. For gas velocity 0.55 m/s, liquid velocity 1.65 m/s Pressure outlet = 0 (pipeline)
3.4.3 Pressure Gradient
One of the key inputs in the design of slurry and slug flow pipelines is pressure gradient (in Pa/m).
Figure 3-5 shows pressure gradients for two-phase slug flow predicted by CFD simulation. As can
be seen, the x axis indicates the liquid superficial velocity, while the y axis indicates the pressure
gradient. Experimental data from studies done by Wang (2015) is used for comparison purposes.
Using a diameter of 0.059 and length 5 m in pipe, and experimental data from Ozbayoglu, M. E.&
Omurlu, C. (2007) Using inner diameter 0.088, outer diameter 0.12 and length 4.57 m in concentric
annuli, the liquid phase is considered as water (density 9982 Kg/𝑚3 and Air (density 1.225 kg/𝑚3).
One set of simulation showed experimental results of liquid velocities ranged from (0.05- 1.16)
0
0.5
1
1.5
2
2.5
3
-1 -0.5 0 0.5 1
Aver
age
Vel
oci
ty (
m/s
)
R/r
Numerical Experimental
47
and constant gas velocity. The simulation results are compared with experimental data as . This
figure shows how the pressure gradients predicted by CFD simulations and obtained from the
experiment are well aligned. This proves the ability of CFD to predict pressure gradients for slug
flows of gas and liquid. Moreover, because the pressure gradient level rose sharply as the liquid
velocity increased, the impact of liquid velocity on pressure gradient can be assumed to be
reasonably high.
Figure 3-5 Pressure gradient obtained from CFD compared to pressure gradients measured from
Experimental data in pipeline and concentric annuli
3.4.4 Profile Slug Volume Fraction in Annular Pipe
Figure 3-6 illustrates slug flow at 0.425 m/s air superficial velocity as well as 0.342 m/s water
superficial velocity through horizontal annular pipe. As shown in Figure 3-6, the air phase is close
to the inner pipe wall. The highest air volume fraction (approximately 37%) can be observed close
to the inner pipe wall, while the lowest air volume fraction (approximately 23%) can be observed
close to the outer pipe wall. This may possibly results due to lift force action in small bubbles.
0
1000
2000
3000
4000
5000
6000
7000
8000
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Pre
ssure
Gra
die
nt
(Pa/
m)
Liquid Velocity(m/s)
Numerical Annular pipe Vg (0.8 m/s) Numerical pipline Vg(7m/s)
Experimental Annular pipe Vg (0.8 m/s) Experimental pipline Vg (7m/s)
48
Additionally, the air volume fraction range for the annulus is around 23 to 37%, as indicated from
the volume fraction profile. Furthermore, the air volume fraction which registers the highest point
is found near the inner pipe wall for the air volume fraction profile.
Figure 3-6 Concentric annular slug flow in horizontal annuli (A)
Figure 3-7 Concentric annular slug flow in horizontal annuli (B)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 0.5 1 1.5 2 2.5
Air
Vo
lum
e F
ract
ion (
%)
Distance of the annular pipe (m)
49
3.4.5 Sensitivity Analysis Comparison between slug flow volume fraction in pipeline and
annular pipe
Table 3.1 Different liquid velocities with constant gas velocity used for simulation
Number Gas
superficial
velocity(m/s)
Liquid
superficial
velocity(m/s)
Pipe
type
1 0.4
0.4
0.3 Pipeline
2 0..3 Annular pipe
3.4.5.1 Slug flow volume fraction in pipeline
In the development phase of multiphase pipelines and related machinery, one of the most important
parameters in both the slug body and gas void fraction is volume fraction. This is because, in
designing the equipment, phase composition and volume fraction must be precisely proportional.
Figure 3-8 shows simulation outcomes of a void fraction for an air-water slug flow regime situated
by the cross-section of a horizontal pipe with length 2 m and dimeter 0.05 m. Figure 3-8 also
illustrates air and water distribution for the horizontal slug flow, with dark blue indicating the air
and red the water, both of which are separated by a line indicating their interface. In the simulation,
water slugs reached as far as the top of the pipe for complete slug regime. The water phase,
however, became unbalanced at the initial wave crest due to sinusoidal perturbation near the inlet.
After that point, sizeable waves were noted, growing increasingly and taking up the entire cross-
section at the pipe end. A lengthy slug could be seen increasing in size near the pipe downstream.
Overall, as the gas superficial velocity increased.
50
Figure 3-8 Sectional of liquid and gas volume fraction evaluation along the pipe length
3.4.5.2 Slug flow volume fraction in annular pipe
As presented in the contour plots, we successfully simulated the slug flow in horizontal annuli pipe
with length of 2 m and outer diameter of 0.04, inure diameter of 0.02 m. Figure 3-9 shows
simulation outcomes of a void fraction for an air-water slug flow, with dark blue indicating the air
and red the water. From the contour plots, we can only visualize slug flow patterns. Figure 3-9
shows the slug portion of the slug unit as well as the accompanying bubble. As can be seen, the
Slug size can only be estimated due to their continuous evolution and change. Figure 3-9 also
shows the positioning of the slug bubbles as well as their distribution at the top and bottom of the
annulus.
Figure 3-9 Sectional of liquid and gas volume fraction evaluation along the annular pipe length
51
3.4.5.3 Slug Flow Volume fraction Analysis with Time
Figure 3-10 and Figure 3-11 show water slugs breaching the upper portion in the pipeline as well
as the annulus pipe, resulting in a complete slug regime. As can be seen, the upper portion is filled
with air with the water slugs portion is filled. In fact, the water portion remains unsteady to the
point of the first wave crest.
Figure 3-10 Slug flow trend at different time lapse indicating time in horizontal pipeline
This is due to sinusoidal perturbation of the inlet. Then, when large waves occurred, the pipe’s
cross-section begins to fill at 3.5s, while the annulus fills at 4.5s
Figure 3-11 Slug flow trend at different time lapse indicating time in horizontal annular pipe
52
3.6 Conclusions
The outcome of the simulation tests clearly showed the ability of the CFD model to demonstrate
the curvature effect resulting from annuli geometry. Because of the effect, the mean axial velocity
profile of a completely turbulent flow within the concentric annuli manifested as asymmetrical and
tilted in the direction of the inner wall. The form of the velocity profile was also impacted by the
inner-to-outer radius ratio. This work attempted to illustrate the flow features of air-water flow
within horizontal pipelines and annular pipe. CFD test simulations were performed with ANSYS
FLUENT software. By applying the VOF approach slug was accurately predicted. As well, the
outcomes of the simulation were validated against earlier experimental results and a reasonably
good agreement was seen for the slug flow pattern. However, flow features; including velocity
profile, pressure, and volume fraction, pertaining to annular and slug flow were investigated. These
findings indicate that total pressure rises as water velocity increases in the annular pipe and
pipeline slug flow, at the same time, the air volume fraction simulations present maximum value
at the cross-section of along the length for both pipes. The outcomes reflect real flow
configurations. Overall, the results from this work could prove helpful in the area of crude oil
transportation, specifically in the development of pipeline infrastructure.
53
Chapter 4
Conclusions and Recommendations for Future Research Work
4.1 Conclusions
In the present study, CFD simulations were performed using ANSYS 16.2 FLUENT software. The
application of the Eulerian model / Reynolds Stress Model (RSM) and the volume of fluid (VOF)
model / Shear-Stress-Transport (SST) model not only showed good slug flow prediction
capabilities but also validated experimental results. As shown, gas-liquid multi-phase slug flows
are highly dependent on flow strength emanating from gas and liquid flow. However, flow
strengths vary and fluctuate according to conditions. Given this situation, it was discovered that
enhanced mesh discretization improved resolution and thus offered better overall outcomes. This
study focused on the theoretical representation of slug flow in both horizontal and annular pipes.
For this, a 3-D CFD / VOF model was employed to serve as the interface for the gas and liquid
and for the tube’s inner fluid flow. Because this theoretical model was shown to succeed at
detecting flow regimes as well as gas volume fraction, CFD should be considered as a viable
approach in predicting tube-based gas/liquid multi-phase flows.
Additionally, the research results could provide a basis for developing crude oil transport pipeline
systems, as using CFD for modeling and flow assurance has been shown to improve the use of
simulations. Hence, although CFD decreases the need for theoretical experimentation, it should
not take the place of experimental analysis. The input data in CFD is often uncertain, prolonging
both the validation and verification stages and having a negative impact on mathematical models.
Thus, prior to applying CFD for modeling, researchers should first take into consideration the
number of computations that will be required to give viable results.
54
4.2 Recommendations for Future Research Work
Further analysis is required with more accuracy which can include choosing different
coefficients and constants to minimize small errors and increase acceptancy of this model
at versatile conditions of operation.
It is expected to find out numerical correlations between different parameters by
conducting further parametric study at distinct phases of fluid flow through pipeline and
annuli.
Complex geometry of pipeline and annuli can be introduced (e.g. bending, inclination etc.).
Elaborate work on Fluid Structure Interaction (FSI) is required focusing the safety and risk
at multiphase slug flow conditions through pipeline and annuli.
55
References
1. Meland, Knut Å. Stabilization of Two- Phase Flow in Risers from Reservoirs - Robust Anti-
Slug Control Strategies. Edited by Sigurd Skogestad., NTNU, 2011.
2. Frank, Thomas. "Numerical simulation of slug flow regime for an air-water two-phase flow
in horizontal pipes." Proceedings of the 11th International Topical Meeting on Nuclear
Reactor Thermal-Hydraulics (NURETH-11), Avignon, France, October. 2005.
3. Crowe, Clayton T., John D. Schwarzkopf, Martin Sommerfeld, and Yutaka Tsuji.
Multiphase flows with droplets and particles. CRC press, Taylor & Francis Group, Boca
Raton, Florida 2011.
4. Van Wachem, B. G. M., and Alf-Erik Almstedt. "Methods for multiphase computational
fluid dynamics." Chemical Engineering Journal 96.1 (2003): 81-98.
5. Stenmark, Elin "On multiphase flow models in ANSYS CFD software." Master thesis,
Chalmers University of Technology, Sweden (2013).
6. Dewangan, S. K., & Sinha, S. L. (2016). Exploring the hole cleaning parameters of
horizontal wellbore using two-phase Eulerian CFD approach. The Journal of Computational
Multiphase Flows, 8(1), 15-39.
7. Soliman, R., & Collier, P. (1990). Pressure drop in slurry lines. Hydrocarbon Processing,
69(11), Hydrocarbon Processing, 1990, Vol.69 (11).
8. Roco, M. C., & Shook, C. A. (1983). Modeling of slurry flow: the effect of particle size.
The Canadian Journal of Chemical Engineering, 61(4), 494-503.
9. Choi, Young-Chan, Park, Tae-Jun, Kim, Jae-Ho, Lee, Jae-Goo, Hong, Jae-Chang, & Kim,
Yong-Goo. (2001). Experimental studies of 1 ton/day coal slurry feed type oxygen blown,
entrained flow gasifier. Korean Journal of Chemical Engineering, 18(4), 493-498.
56
10. Huilin, Gidaspow, & Bouillard. (2002). Chaotic behavior of local temperature fluctuations
in a laboratory-scale circulating fluidized bed. Powder Technology, 123(1), 59-68.
11. Brandstaetter, W., Ragab, A., & Shalaby, S. (2007). Modeling of Two-Phase Flow, and Slug
Flow Characteristics in Horizontal/Inclined Pipelines using CFD. SPE Romanian Section
Conference.
12. Frank, T. (2005, October). Numerical simulation of slug flow regime for an air-water two-
phase flow in horizontal pipes. In Proceedings of the 11th International Topical Meeting on
Nuclear Reactor Thermal-Hydraulics (NURETH-11), Avignon, France, October (pp. 2-6).
13. Al-Lababidi, S., Addali, A., Yeung, H., Mba, D., & Khan, F. (2009). Gas void fraction
measurement in two-phase gas/liquid slug flow using acoustic emission technology. Journal
of Vibration and Acoustics, 131(6), 064501.
14. Czapp, M., Utschick, M., Rutzmoser, J., & Sattelmayer, T. (2012, July). Investigations on
slug flow in a horizontal pipe using stereoscopic particle image velocimetry and CFD
simulation with volume of fluid method. In 2012 20th International Conference on Nuclear
Engineering and the ASME 2012 Power Conference (pp. 477-486). American Society of
Mechanical Engineers.
15. Razavi, S. Y., & Namin, M. M. (2011). Numerical model of slug development on horizontal
two-phase flow. In Proceedings of the International Conference on Recent Trends in
Transportation, Environmental and Civil Engineering (pp. 14-15).
16. Al-Hashimy, Z. I., Al-Kayiem, H. H., Time, R. W., & Kadhim, Z. K. (2016). Numerical
Characterisation of Slug Flow in Horizontal Air/water Pipe Flow. International Journal of
Computational Methods and Experimental Measurements, 4(2), 114-130.
57
17. Taitel, Y., & Dukler, A. E. (1976). A model for predicting flow regime transitions in
horizontal and near horizontal gas-liquid flow. AIChE Journal, 22(1), 47-55.
18. Ghorai, S., & Nigam, K. D. P. (2006). CFD modeling of flow profiles and interfacial
phenomena in two-phase flow in pipes. Chemical Engineering and Processing: Process
Intensification, 45(1), 55-65.
19. Tan ping, FU xingjun et al. Study on Vibration of Conveying Water Piping System in Power
Plant, Turbine Technology, Vol .46, No.2, 2004.
20. Chica, L., Pascali, R., Jukes, P., Ozturk, B., Gamino, M., & Smith, K. (2012, July). Detailed
FSI Analysis Methodology for Subsea Piping Components. In ASME 2012 31st
International Conference on Ocean, Offshore and Arctic Engineering (pp. 829-839).
American Society of Mechanical Engineers.
21. Anderson, T. B., & Jackson, R. (1967). Fluid mechanical description of fluidized beds.
Equations of
22. Bowen, R. M. (1976). Theory of mixtures. Continuum physics, 3(Pt I)
23. Alder, B. J., & Wainwright, T. E. (1960). Studies in molecular dynamics. II. Behavior of a
small number of elastic spheres. The Journal of Chemical Physics, 33(5), 1439-1451.
24. Chorin, A. J. (1968). Numerical solution of the Navier-Stokes equations. Mathematics of
computation, 22(104), 745-762.
25. Launder, B. E. (1989). Second‐moment closure and its use in modelling turbulent industrial
flows. International Journal for Numerical Methods in Fluids, 9(8), 963-985.
26. Chica, L. (2014). FSI study of internal multiphase flow in subsea piping components
(Doctoral dissertation, University of Houston).
58
27. Nouri, J. M., Umur, H., & Whitelaw, J. H. (1993). Flow of Newtonian and non-Newtonian
fluids in concentric and eccentric annuli. Journal of Fluid Mechanics, 253, 617-641.
28. Chung, S. Y., Rhee, G. H., & Sung, H. J. (2001). Direct Numerical Simulation of Turbulent
Concentric Annular Pipe Flow. In TSFP Digital Library Online. Begel House Inc.
29. Al-Hashimy, Z. I., Al-Kayiem, H. H., Kadhim, Z. K., & Mohmmed, A. O. (2015). Numerical
simulation and pressure drop prediction of slug flow in oil/gas pipelines. Computational
Methods in Multiphase Flow VIII, 89, 57.
30. Kristiansen, O. (2004). Experiments on the transition from stratified to slug flow in
multiphase pipe flow. PhD thesis, Norwegian University of Science and Technology
(NTNU) Trondheim.
31. Orell, A. (2005). Experimental validation of a simple model for gas–liquid slug flow in
horizontal pipes. Chemical Engineering Science, 60(5), 1371-1381.
32. Jepson, W. P., & Taylor, R. E. (1993). Slug flow and its transitions in large-diameter
horizontal pipes. International journal of multiphase flow, 19(3), 411-420.
33. Ghorai, S., & Nigam, K. D. P. (2006). CFD modeling of flow profiles and interfacial
phenomena in two-phase flow in pipes. Chemical Engineering and Processing: Process
Intensification, 45(1), 55-65.
34. Thome, J. R. (2004). Engineering data book III. Wolverine Tube Inc, 2010.
35. De Henau, V., & Raithby, G. D. (1995). A transient two-fluid model for the simulation of
slug flow in pipelines—I. Theory. International Journal of Multiphase Flow, 21(3), 335-349.
36. Ghosh, S., Das, G., & Das, P. K. (2010). Simulation of core annular down flow through
CFD—A comprehensive study. Chemical Engineering and Processing: Process
Intensification, 49(11), 1222-1228.
59
37. Kaushik, V. V. R., Ghosh, S., Das, G., & Das, P. K. (2012). CFD simulation of core annular
flow through sudden contraction and expansion. Journal of Petroleum Science and
Engineering, 86, 153-164.
38. Ghosh, S., Das, G., & Das, P. K. (2011). Simulation of core annular in return bends—a
comprehensive CFD study. Chemical Engineering Research and Design, 89(11), 2244-
2253.
39. Fluent, A. N. S. Y. S. "Ansys Fluent Theory Guide." ANSYS Inc., USA (2011).
40. Wen, J. X., Le Fur, P., Jie, H., & Vendra, C. M. R. (2016). Further development and
validation of C𝑜2 FOAM for the atmospheric dispersion of accidental releases from carbon
dioxide pipelines. International Journal of Greenhouse Gas Control, 52, 293-304
41. Alder, B. J., & Wainwright, T. E. (1960). Studies in molecular dynamics. II. Behavior of a
small number of elastic spheres. The Journal of Chemical Physics, 33(5), 1439-1451.
42. Anderson, T. B., & Jackson, R. (1967). Fluid mechanical description of fluidized beds.
Equations of motion. Industrial & Engineering Chemistry Fundamentals, 6(4), 527-539.
43. Menter, F. R. (1994). Two-equation eddy-viscosity turbulence models for engineering
applications. AIAA journal, 32(8), 1598-1605.
44. Wilcox, D. C. (1998). Turbulence modeling for CFD (Vol. 2, pp. 103-217). La Canada, CA:
DCW industries
45. Walters, D. K., & Cokljat, D. (2008). A three-equation eddy-viscosity model for Reynolds-
averaged Navier–Stokes simulations of transitional flow. Journal of fluids engineering,
130(12), 121-401.
46. Ozbayoglu, M. E., & Omurlu, C. (2007). Modelling of two-phase flow through concentric
annuli. Petroleum science and technology, 25(8), 1027-1040.
60
47. Nouri, J. M., & Whitelaw, J. H. (1994). Flow of Newtonian and non-Newtonian fluids in a
concentric annulus with rotation of the inner cylinder. Transactions-American Society of
Mechanical Engineers Journal of Fluids Engineering, 116, 821-821.
48. Lewis, S., Fu, W. L., & Kojasoy, G. (2002). Internal flow structure description of slug flow-
pattern in a horizontal pipe. International journal of heat and mass transfer, 45(19), 3897-
3910.
49. H. Hadia, R. A. Sultan, M. A. Rahman, J. Shirokoff, S. Zendehboudi, A Computational Fluid
Dynamics Study of Two-Phase Slurry and Slug Flow in Horizontal Pipelines, 3rd Workshop
and Symposium on Safety and Integrity Management of Operations in Harsh Environments
(C-RISE3).
50. H. Hadia, J. Shirokoff, M.A. Rahman, R. A. Sultan, S. Zendehboudi, Analysis of Two Phase
Slug Flow through Annular Pipeline, Qatar Conference on CFD, (2017).
51. I. A. Adeoti, R. A Sultan H. Hadia, M. A. Rahman, J. Shirokoff, A. Maynard, and M. Addy,
CFD Approach to Analyze Slug Flow Induced Fluid Structure Interaction (FSI) through
Pipeline and Annuli, Journal of Computational Science (2017) submitted.